MR volume estimation of subcortical brain lesions and ventricular cerebrospinal fluid: a simple and accurate stereologic method.

PURPOSE To describe an MR imaging quantification method for estimation of total volumes of both white and gray matter subcortical lesions and ventricular cerebrospinal fluid (CSF) in the living human brain, and to determine the method's reliability. METHODS In 12 subjects, total subcortical lesion and ventricular CSF volumes were estimated using systematic sampling. Systematic sampling was performed on equidistant MR sections using a counting grid with systematically ordered intersection points. The grid was randomly positioned on each consecutive MR section. Each grid intersection point hitting the structure of interest represents a fixed known volume dependent on grid intersection point distance and the sum of the section thickness and section gap. RESULTS Total volume estimation of subcortical lesion and ventricular CSF takes 15 and 5 minutes per subject, respectively. Coefficients of error of the individual volume estimates ranged from .01 to .13 and are negligible to the coefficients of the group mean (range, .70 to .89). For subcortical lesion volume, the random intraobserver error yielded .04 and for ventricular CSF .02; the random interobserver error amounted to .11 and .04, respectively; and the systematic interobserver error was .15 and .04, respectively. CONCLUSION The method described here for subcortical lesion and ventricular CSF volume estimation is accurate, reliable, valid, and fast.